YIELD DESIGN: A SURVEY OF THE THEORY

J.

Salen~on

Ecole Polytechnique, Palaiseau, France

ABSTRACT
The theory of Yield Design is based upon the obvious necessary condition for the
stability of a structure that the equilibrium of that structure and the resistance of its
constituents should be compatible. The static approach of the yield design theory
proceeds directly from this condition, leading to lower estimates of the extreme loads.
The kinematic approach is derived by dualizing the static approach through the
principle virtual work, thus ensuring full mechanical consistency. The treatment of a
classical example illustrates these arguments. Present and possibly future domains of
practical applications of the theory are reviewed, including the full adequacy between
the Yield Design Theory and the Ultimate Limit State Design concept of safety.

G. S. Landriani et al. (eds.), Evaluation of Global Bearing Capacities of Structures

Springer-Verlag Wien 1993

J.

Salen~on

1.- SOME HISTORICAL LANDMARKS

The concept of yield (or limit) design is most popular among civil engineers who
have made an extensive use of it for centuries when designing structures or
earthworks. As a matter of fact, one usually refers to Galilee's study [1] of the
cantilever beam as the first written explicitation of the concept : the maximum load the
beam may withstand is derived from the only knowledge of the strength of its
constituents, namely the fibers (Fig.1).

Figure 1 - Galileo's analysis of the cantilever beam.

Coulomb's celebrated memoir [2] appears then as the reference where such
problems as the compression of a column, the stability of a retaining wall or of a
masonry arch, etc., are considered from the same point of view combining the
equilibrium of the structure and the resistance of its constituents. Papers by Heyman
[3-7] and by Delbecq [8, 9] may help getting a comprehensive view of Coulomb's
analysis and of the subsequent works (e.g. Mery [1 0]. Durand-Ciaye [11, 12], ...)
concerning masonry works.
As regards soil mechanics, the original Coulomb analysis has been followed by
others in the same "spirit" : the general equilibrium, in terms of resulting moments
and resulting forces, of a bulk of soil, defined in the considered earthwork by one or a
few parameters, is checked under the condition that the soil cannot withstand stresses
outside its strength criterion usually defined as a Coulomb criterion. Among many, one
must quote the Culmann method [13]. and Fellenius' famous analysis for slope stability
where the bulks of soil whose equilibrium is checked are limited by circular lines.
When considering such circular lines in the case of a frictional soil, the reasoning can

Yield Design: A Survey of the Theory

only be carried out by introducing complementary assumptions as in the slices method

for instance [14-17). Rendulic [18] introduced the use of logspirals instead of
circular lines in that case, which removes the difficulty and preserves the full
significance of the analysis as it will be shown later on. Analogous analyses have been
performed in order to investigate the stability of surface foundations.
For what concerns the stability of surface foundations and the determination of
their bearing capacity, the limit

equilibrium methods (as classified by Chen [19) for

instance) have also been used, especially in the case of plane problems : the equations
were first established by Massau [20) and KOtter [21, 22], and important
developments may be found in the famous books by Sokolovski [23-25] and Berezancew
[26). Since the limit equilibrium equations for plane problems are homologous to the
equations appearing in the study of plane strain problems for perfectly
materials obeying Tresca's yield

plastic

criterion with associated flow rule (27], they are

often associated with the concept of plasticity : in other words they are improperly said
to rely on the assumption of a rigid perfectly plastic soil obeying Coulomb's criterion
with associated flow rule. As a matter of fact, it should be clear to everybody that limit
equilibrium methods are but another technique to perform the reasoning of the yield
design theory, and that they do not require any more assumption about the constitutive
law of the considered soil than the data of its strength criterion.
Another classical domain of application for the yield design theory concerns the
analysis of metallic plates and reinforced concrete slabs, with the use of Johansen's
criterion and yield line theory [28, 29] and other methods, as explained for instance in
Save and Massonnet (30).
Recently the combination of the homogeneization theory with the yield design
theory has opened the access to new types of investigations and has already led to
interesting results such as the determination of the "macroscopic" homogenized
strength criterion of a composite based upon the data of the strength criteria of its
constituent materials, including the interfaces between them, which is to be introduced
in the homogenized yield design analysis of a structure made of that composite material.
From the practical point of view the domain of relevance ranges from mechanical
engineering with, for instance, the "classical" long fiber/epoxy matrix composites
[31, 32) to civil engineering where soils reinforced by inclusions such as nails,

J. Salenc,:on

geotextiles and geogrids, or metallic strips, etc., are now being extensively used

[33-35).
The purpose of the presentation of the yield design theory to be given hereafter is to
cover those different types of applications, and others which have not been listed to
make short. Starting from the simple example of a structure, in order to introduce and
discuss the first fundamental concepts, it will then be developed within the framework
of classical continuum mechanics with the help of a few basic notions in convex
analysis.
Through its generality it is hoped to show the unity between existing methods in the
various domains of applications of the yield design theory, together with their
efficiency, and to arouse the interest for devising new developments which may take
advantage of the everyday increasing possibilities offered by numerical techniques.
2.- AN INTRODUCTION TO THE YIELD DESIGN APPROACH.

2.1.- Definition of the problem under consideration

The square truss ABCD shown in Fig. 2 is made of six bars with pivot hinges in A,
8, C and D, and no connection between bars AC and BD. This structure is submitted to an
external force acting on joint C, in the direction of AC with intensity Q, while joint A is
fixed. No other force is applied to the truss and Q appears naturally as the load
parameter for the problem.
The internal efforts in the bars obviously reduce to tensile or compressive forces,
namely N1 , N2 ,... , N6 in bars AB, BC, CB, DA, AC and BD respectively.

Figure 2- One parameter loading of a simple structure.

Yield Design: A Survey of the Theory

It is assumed that the six bars of the truss exhibit the same resistance in tension
and in compression, equal to L, so that condition (2.1) has to be satisfied :

vi

I Ni I ~ L

= 1' 2, ... , 6.

( 2. 1 )

The question to be answered is to determine the maximum and minimum

values of parameter Q the structure can withstand under the strength conditions (2.1)
imposed to the bars.

2.2,._ Compatibility between equilibrium and resistance.

The equilibrium equations for the structure in Figure 2 are written

N1

N2

N3

N5 + N1'1/2

= Q ,

N6 + N1'1/2

N4,
(2.2)

A necessary condition for that structure to withstand load Q under the strength
conditions (2.1) is therefore that Eq. (2.2) be mathematically compatible with lneq.
(2.1 ). It follows that the intensity Q of the applied force should not exceed the
conditions :

101

(2.3)

~2L

This leads to the following statements :

if the applied load Q is such that

101

> 2L, it is sure that the structure cannot be

in equilibrium together with the strength conditions being satisfied at the same
time;
if the applied load Q is such that

101

2L, it is possible that the structure be in

equilibrium and the strength conditions be satisfied at the same time.

Briefly speaking, introducing the termes "stable" and "unstable" to characterize
those circumstances :

101 > 2L :

the structure cannot be stable under Q ,

it is surely unstable under Q ;

101

(2.4)

2L : the structure may be stable under Q,

it is potentially stable under Q.

The loads Q+ = 2L and

a-

= -

in the loading mode defined in Fig. 2.

2L will be called the extreme loads of the structure

J.

Salen~on

2.3. Comments.
Statement (2.4) deserves many comments, most of which will remain relevant in
the general case.

2.3.1.
The analysis has been performed on the given geometry of the structure : no
geometry changes are taken into account.

2.3.2.
The only data required for the analysis given in Section (2.2) are the strength
conditions imposed on the bars. It follows that results (2.4) hold whatever the initial
internal forces in the structure for 0 = 0, whatever the loading path, whatever the
constitutive equations for the bars provided they are consistent with the strength
conditions (2.1) and with the preceding comment regarding the geometry where the
equations are written.

2.3.3.
Statement (2.4) is but a partial answer to the original question asked in Section
(2.1) since it only offers a guarantee of instability (!) when
presumption of stability when

101

101

> 2L and a

s 2L.

It should be understood that such a conclusion is the maximum one can logically
derive starting from the only available data (2.1) on the resistance of the constituent
elements of the structure imposed as a constraint. In order to be able to assert the
stability of the structure under a given load 0 it would be necessary that
complementary information regarding
- the mechanical behaviour of the constituent elements, through their constitutive
equations for instance,
- the initial state of self-equilibrated interior forces,
- the loading history followed to reach the actual load 0 from the unloaded initial state
be available.
As a matter of fact it can be easily perceived that the condition for the structure to
withstand an extreme load is that the strength capacities required to equilibrate that
load can be actually mobilized simultaneously in the concerned elements, at the end of
the concerned loading path, starting from the given initial state of self equilibrated
interior forces, and with the given constitutive equations.

Yield Design: A Survey of the Theory

Fig. 4 recalls an example from [36] where the response of the structure
introduced in Fig.2 is studied for a monotous increasing loading path, starting from the
zero self equilibrated initial state of interior forces, under two different assumptions
for the constitutive equations of the elements (Fig.3) :
a) the bars are elastic and perfectly plastic in tension and in compression,
b) the bars are elastic and perfectly plastic in compression, elastic and brittle in
tension.

b)

Figure 3 - Elastic perfectly plastic (a), and elastic brinle/perfectly plastic (b)
behaviours for the bars in Fig. 2.
Although the latter assumption might seem somewhat schematic, the results shown
in Fig. 4 point out that in the first case the structure will actually withstand any load
between

a and a+

the level

a = L"2.
2L

whereas in the second case the structure will break when

a reaches

L./2

b)

Figure 4 - Response of the truss in Fig. 2 with the constitutive equations

shown in Fig. 3 for the bars.

J. Salenyon

The preeminent role of ductility in the mechanical behaviour of the bars is quite
apparent here. Generally speaking such ideal behaviours as shown in Fig. 3 will
scarcely be encountered when dealing with practical applications. Nevertheless it is
clear that the idea expressed by Jewell [37] that the deformations necessary to fully
mobilize the required resistance of the elements, in the structure, along the considered
loading path, must be "compatible" with each other must be retained.

3.- PRINCIPLES OF THE THEORY OF YIELD DESIGN: STATIC APPROACH

3. 1.

Notations

The theory will be presented within the frame of classical continuum mechanics.
The following notations will be used :

n : volume of the system

under consideration ;

an

:boundary of

n;

: vector defining the position of the current point in

: Cauchy stress field ;

: velocity field ; U : virtual velocity field ;

: strain rate field derived from U , (St from U) ;

n;

[ ] : symbol for the jump accross surface :t following unit normal n ;

: symbol for the contracted product (e.g.

a.. n

<Jij nj and T.U = Ti Ui (1)) ;

: symbol for the twice contracted product (e.g. a.: .d. = <Jij dji).

3.2. Principle of virtual work (powers). Loading parameters.

The theory of yield design is most conveniently formulated in the case when the
loading mode of the system depends on a finite number of loading parameters, which is
always the case in practical applications. The definition of such loading parameters is
given in [36]. It may be shortly expressed as follows, through the principle of virtual
work in statics.

(1) Orthonormal cartesian coordinates.

Yield Design: A Survey of the Theory

For any stress field

, statically admissible for the problem in the loading mode,

for any virtual velocity field

U,

kinematically admissible for the problem in the

loading mode,
the following equation holds

a(x) : i(x) dn

+f. [

U (x )] . a.(x) . n(x) dl:

= Qj(a.). qi (U) = Q(a.) . cHU)

( 3. 1)

where the applications

~ ~
.......

R"

(3.2)

= q(U) E

(3.3)

Q = Q(a)
""'

~ q

"""

are linear.
The Oj. components of the n-dimension vector Q, are the loading parameters of the
problem, Q being called the load. The qj. components of the n-dimension vector

q, are

the associated kinematic parameters.

It is recalled that the statically admissible stress fields a. in Eq. (3.1) are
piecewise continuous with continuous derivatives, and satisfy the equilibrium equations
on

n and the boundary conditions on the stresses on an.

The kinematically admissible

virtual velocity field are piecewise continuous with continuous derivative on

satisfy the boundary conditions on the velocity on

and

an ; it must be emphasized that no

restriction are imposed on the velocity discontinuities [U).

Classical examples of loading parameters are given in [36) ; de Buhan [38],
dealing with systems made of composite materials through the homogenization theory,
gives interesting developments on the concept of loading parameters regarding both the
structure itself and the unit representative cell.

3.3.- Statement of the problem.

The general formulation of the yield design problem is similar to that given in
Section (2.1) in the particular case of the truss in Fig.2.
The system under consideration is geometrically defined through

n.

It is loaded

according to a loading mode depending on n parameters Oj The strength condition,

homologous to lneq. (2.1) must then be defined. Obviously, such a condition will now
refer to the value of the stress tensor ~(x) at each point x in

n.

J. Salencon

10

The strength domain G(x) defining the admissible stress states at point x is given
as the only data to characterize the constituent material mechanically. It depends
explicitly on x since the material needs not be homogeneous. The strength condition will
be written :
'v'XEO

a.(x)

(3.4)

G(x) c R

As a rule G(x) exhibits the following properties

!l(X) =

g_ E G(x) ,

G(x) is star-shaped with respect to Q..

As a matter of fact G(x) is usually convex (that implies the latter property),
which will assumed from now on.
The question is then to decide whether the system can or cannot withstand a given
value of load Q, under the constraint of the strength condition (3.4).

3.4. Potentially safe loads, extreme loads.

The answer follows from the same considerations as in chapter 2.
For the system to withstand load Q under the strength condition (3.4) it is
necessary that :
3 Q.

statically admissible for the problem

in the loading mode with load Q ,
complying with the strength condition (3.4).

(3.5)

Under such a load Q, the system will be said potentially stable, and the load itself
will be called potentially safe.
The set generated by all potentially safe loads will be denoted K :
Q

K c R" ~

3a.

S.A. equilibrating Q
a.(x)

G(x)

vx E n

(3.6)

stress field .a. is associated with load Q through the linear application (3.2).
The properties of K are derived immediately from definition (3.6) and the
linearity of application (3.2). It comes out that :
Q = OE K '
K is convex ,

(3.7)

11

Yield Design: A Survey of the Theory

what proves that the properties of the G's at the level of the constituent material are
transferred to K at the level of the system.
The loads on the convex boundary of K are called the extreme loads of the system,
which recalls that any load out of K is certainly unsafe and will induce instability of the
system.
extreme loads

Figure 5 - Domain K :potentially safe loads, extreme loads.

3.5.-

Comments.

Comments homologous to those expressed in section (2.3) should be made. The

answer illustrated in Fig.S is but a partial one to the original question : domain K offers
no guarantee of stability for the system, which would require the data of the initial
state, of the loading history, and of the mechanical behaviour of the constituent
material.
It is hoped that the above given presentation made it clear that the definition of
domain K is only derived from the

Compatibility between

/ equilibrium (of the system)

\ resistance (of the material)

(3.8)

no other data being available. Therefore, what may appear a shortcoming in the result
can also be considered positive : domain K is independent of the initial state, of the
loading history of the system and of the mechanical behaviour of the constituent
material apart from the strength condition (3.4). Since this information may be either
totally or partially missing in some cases, or improperly assessed in others, it is
important that such a general result as K be available.

12

J.

Salen~on

Anyhow the question of the relevance of K for practical applications cannot be

disregarded.
From the theoretical point of view it has been proved [39] (ct. (40]) that a
system whose constituent material is elastic and perfectly plastic and obeys the
maximum plastic work principle will withstand any load Q inside K (which generalizes
the result in Fig. 4a). In that case the relevance of K is established.
This theoretical situation is seldom encountered in practice and the relevance of K
must be assessed by experiments and experience. The general idea already expressed in
para (2.3.2.), that the deformations needed to fully mobilize the required strengths in
the element of the system must be compatible, should be retained. AUention should also
be paid to the initial geometrical assumptions and the incidence of geometry changes
should be considered.
As an example it appears that in Soil Mechanics, for usual stability analysis
purposes, the reasoning based upon the compatibility stated in (3.8) which is the only
and very argument of the Yield Design Theory, has proved to be relevant in most cases,
but is not valid when dealing for instance with the stability of deep tunnels or deep
cavities.
Such validations of the theory support the determination of K and, more precisely,
that of the exteme loads.

3.6. Static approach from Inside.

The construction of domain K comes directly from definition (3.6) : any
stress-field .a.. statically admissible in the loading mode and complying with the
strength condition in n produces a potentially safe load Q = Q(a) through application
(3.2). In other words :
a S.A. in the mode

.a.(x) e G(x) ,

vx

(3.2)

Q(a)

K .

(3.9)

Drawing the convex envelope of such Q gives an interior approach of K and "lower
bounds" for the extreme loads (Fig. 6).
The exact determination of K requires exploring the whole set of

a satisfying

(3.9).

This is most often impossible from the practical point of view, except for the case of

13

Yield Design: A Survey of the Theory

simple structures (as in Section 2). It follows that one will usually be satisfied with
the interior approach derived through the implementation of construction (3.9) with a
reduced number of stress-fields chosen for an efficient balance between simplicity in
the procedure and quality of the obtained results.

Figure 6 - Static approach from inside.

4.- AN EXAMPLE OF THE YIELD DESIGN STATIC APPROACHES.

4.1.- Stability analysis of a vertical cut.

The stability of the homogeneous vertical cut represented in Fig. 7 is investigated.
The load is defined by the gravity forces acting in the bulk of soil, the corresponding
loading parameter being the voluminal weight usually denoted y in Soil Mechanics.
The strength condition of the soil is characterized through Coulomb's isotropic
strength criterion with cohesion C and friction angle cp, written in the form
f (sz..(x)) = . Sup
I,)=

1, 2, 3

{a; (1 + sinc:p)- ai (1- sinc:p)- 2 C cosc:p} ~ 0

where aj denote the principal values of a.(x).

z
h

Figure 7 - Stability of a homogeneous vertical cut under its own weight.

(4.1)

J.

14

Salen~on

4.2.- Static approach from inside.

A stress-field for the static approach from inside was given by Drucker and Prager

[41] and proved to comply with lneq. (4.1) as long as

(yh/C) :s; 2 tan(7tl4 + <p/2) ;

( 4. 2)

it is presented in Fig.B.
It follows that the value 2 tan(7tl4 + <p/2) is a lower bound for the extreme value of
the non-dimensional parameter (yh/C) which controls the stability of the cut :
(yh/C) +

2 tan( 7tl4 + <p/2) .

(4.3)

---'---+-------------..
t)~u. Yy

71-a"" Yy
ayy Yy

Figure 8 - Static approach for the problem of the vertical cut [41].

4.3.- Static approach from outside.

Still starting from definition (3.6) of K it will now be shown that the static
reasoning

may be performed in a different way in order to provide an upper bound for

(yh/C)+ .
Let the bulk of soil be virtually separated into two parts (1) and (2) by means of
an arbitrary plane (P) passing ghrough the bottom line of the cut as drawn in Fig.9.
The compatibility stated in (3.8) implies that, for any such plane (P), the global
equilibrium -i.e. from the statics of rigid bodies point of view- of volume (1) or
volume (2) be possible under the action of the vertical forces due to gravity (bulk
weight of the soil) and of the resisting forces developed by the material.
The cut being considered of infinite length along Oz, the resisting forces developed
on the sections parallel to OAB at both ends at infinity are negligible compared with

Yield Design: A Survey of the Theory

15

those along the section by plane (P) : the analysis can therefore be restricted to
checking the stability of a slice with thickness D along Oz as sketched in Fig.9 assuming
no resisting forces on OAB and O'A'B'.
Considering the equilibrium of volume (1) submitted to gravity forces acting
throughout the volume and to normal and tangential stresses acting on the surface
ABB'A', the following equations shall be satisfied :
2
yh 0 tana sina = -

(4.4)

odS

2
yh 0 sina t
2

=- f

(4.5)

t dS

fofBA'

where o and t (a vector) are the normal and tangential components of the stress vector
acting at each point of ABB'A', tensile stresses are counted positive, and t is the unit
vector along BA.

Figure 9- Global equilibrium analysis of volume (1).

Coulomb's strength condition (4.1) implies that whatever the point in ABB'A', o
and t must comply with the inequality
1t 1 ::;

C - o tan<p

(4.6)

J.

16

Salen~on

It follows from Eq. (4.5) and (4.6) that

2

yh D sina
2

sf

ltl

dS

N!iJ/11

sf

(C-atan<p)dS

(4.7)

-A'

then through Eq. (4.4)

yh/C

2 cos<p I sin a cos (a + <p)

(4.8)

establishing a necessary condition for the stability of the cut, to be satisfied whatever
the plane (P).
The minimum of the second hand of lneq. (4.8) with respect to a which provides
the strongest necessary condition on yh/C, is obtained for a = 7tl4 - <p/2 which yields the
inequality
yh/C s 4 tan(x/4 + <p/2)

(4.9)

with the conclusion that any cut for which (yh/C) is greater than 4 tan(7tl4 + <p/2) will

be unstable ; thence :
(yh/C)+ s 4 tan(x/4 + <p/2) .

4.4.-

(4.1 0)

Comments.

It must be emphasized that this latter result is an upper bound for the extreme
value (yh/C)+. This is often confusing since, being of the "static" type, one would
readily expect this approach to be an interior one and to yield a lower bound. It is
therefore important to point out where the difference comes from, which may be
expressed as follows.
Compatibility between equilibrium (of the system) and resistance (of the
material), instead of being thoroughly checked as in (3.6) or (3.9) is only investigated
partially, regarding the resulting force in the global equilibrium of an arbitrary
triangular shaped volume. Complying with this partial requirements does not
automatically ensure that conditions (3.9) will be satisfied and, therefore, does not
provide an interior approach of K, i.e. a lower bound for (yh/C)+. Conversely, not
complying with this requirement implies that conditions (3.9) cannot be satisfied
simultaneously, which yields an exterior approach of K, i.e. an upper bound for
(yh/C)+.
The importance of getting such an exterior approach of K is evident from what has
been said about the practical application of the static approach form inside in Section

Yield Design: A Survey of the Theory

17

(3.6). It is also apparent on the above given example that such a static approach from

outside
- requires great care in its implementation,
- remains rather crude : it would certainly be worth carrying on thoroughly with
the same idea, within the framework of continuum mechanics, that means
proving incompatibility between equilibrium and resistance without restricting
the analysis to "global" considerations.
The following chapter will present a method wich meets this latter goal together
with being systematic as regards its implementation.
5.- PRINCIPLES OF THE THEORY OF YIELD DESIGN: KINEMATIC APPROACH.

5.1. Fundamental statement of the kinematic approach.

Starting from definition (3.6) of domain K, Eq. (3.1) expressing the principle of
virtual work may be written for any potentially safe load Q, considering one stress
field

a.

referred to in (3.6) and any k.a. virtual velocity field

g_(x} : d_(x} dn +

I[

U:

U (x}]. g_(x} . n (x} dl:

,..

,..

( 5.1}

Q(g). q(U) = Pe(Q, U)

The following "x functions" will be introduced :

x (x, d_(x}} = Sup {st'(x}: d.(x}

x (x, n(x}, [U(x}]}

g_'(x} e G(x}}

Sup {[U (x}]. g_'(x}. n(x}

a.'(x} e G(x}}

(5.2}

(5.3}

Since stress field a. in Eq. (5.1) complies with strength condition (3.4) in n, it
follows from Eq. (5.1 to 5.3) that :
,..
'V Q e K , 'V U k.a ,
Pe(Q, U} = Q. tl{U}

~{

(5.4}

x (x, d.(x}} dn + J.x (x, n(x}, [U(x}]} dl:

The second hand of lneq. (5.4} will be denoted

Pmr(U) :

J.

18
and called the maximum resisting power in virtual velocity field

Salen~on

U.

The obtained result will thus be written in the form :

.....

'<:/ U k.a

K c

5.2.

"tc

{Q. cHU) - Pmr(U) ~

0}

(5.6)

functions".

5.2.1.- Mechanical significance of the tc functions.

Statement (5.6) clearly relies on the introduction of the x functions as defined by

Eq. (5.2). The significance of these functions must now be investigated .

...

Eq. (5.2) shows that, given a symmetric tensor .d.(x) which may be interpreted as
a virtual strain rate at point x, x(x, .d..(x)) represents the maximum value of the
work ~(x) : .d.(x) which may be developed by any stress tensor ~(x) complying with

...

...

strength condition (3.4) at point x. Function x(x, d..(x)) therefore appears as the
maximum resisting power density in strain rate d..(x) under strength condition (3.4)

...

defined by G(x).
From the mathematical point of view x(x, ) is the support function of convex
G(x) (42) and it may be proved that the data of x(x, ) is equivalent to that of G(x)
(through a strength criterion for instance), which means that x(x, ) carries, in a
dual formulation, all information contained in G(x).
The origin of the word "resisting" in the terminology is evident from Eq. (3.1)
written as a balance between the power of the external forces and the resisting power
(opposite to the power of the internal forces).
The same interpretation as given above holds for x(x, n(x), [U(x)]) since it may
be easily verified that
x (x, n(x), [U(x)])

= .Lx
2

(x, n(x) [U(x)] + [U(x)) n(x))

(5.7)

It is worth pointing out that the fundamental idea of the kinematic approach, that is
defining the strength domain by duality through the x function, is apparent in a paper
by Prager [43].
5.2.2.- Calculation of the tc functions.

The values of the x functions are obtained from Eq. (5.2) and (5.3). Considering
Eq. (5.7), only the case of function x(x, d(x)) will be investigated.

19

Yield Design: A Survey of the Theory

Fig.1 0, schematically presents the two typical circumstances which are

encountered when calculating x(x, d(x)).

1o_ If convex G(x) is bounded in any direction of Rs, then :
,..

x (x, d_(x))

= 2..

,..

(x) : d,.(x)

(5.8)

where ~*(x) is any stress tensor on the boundary of G(x) where the outward normal is
colinear to d_(x). It follows that, in that case, x(x, d.(x)) is finite whatever d_(x).
2- If convex G(x) is not bounded in any direction, let (I) be the convex cone of the
directions in which it extends to infinity in R6. Then :
if .d.(x) belongs to the convex cone orthogonal to (I), x(x, i(x)) has a finite
value

,..

1t

,..

(x, d_(x)) = 2.. (x) : d_(x)

(5.9)

where a.*(x) is any stress tensor on the boundary of G(x) where the outward normal is

colinear to d.(x), as in the preceding case ;

if .d.(x) does not belong to that cone, 7t(X, d.(x)) is infinite

x(x, d_(x)

oo

(5.10)

Figure 10- Calculation ofx(x, d(x)).

Within the framework of classical continuum mechanics, strength domains under
general three dimensional stress condition are usually of the 2nd type presented in Fig.
10 (e.g. Tresca's, von Mises', Coulomb's, ... strength criteria), and the corresponding x
functions take finite or infinite values according to the given d.(x). It is obvious that
only in the case when G(x) is unbounded in any direction, that is the ideal rigid
material, will n(x, d.(x)) be infinite whatever d.(x).
Strength domains for plane stress loading of the material are bounded in any
direction, as are the strength domains referring to bending moments in the yield design

20
of plates and the strength domains for bending moments, normal and shear forces, etc.
in the yield design of beams.
A comprehensive, but of course non exhaustive, list of strength criteria and
corresponding 1t functions is given in [36]. A shorter list appears in Section 5.4.

5.2.3.- Properties of the rr functions.

The following properties are evident from Eq. (5.2) and the assumptions on G(x) :
1t(x, ) is non negative,

1t(X, ) is positively homogeneous of degree 1,

1t(X, ) is convex.

5.3. The kinematic approach from outside.

5.3.1.- Exterior approach.

Statement (5.6) provides a powerful exterior approach of K.

~

For any virtual k.a. velocity field U the maximum resisting power Pmr(U) is
computed referring to the relevant expressions of the x functions, while the power of
A

the external forces Pe(O, U)

=Q

. q(U) is a linear form in Q where q(U) is known. It

follows that domain K is included in the halfspace of Rn defined by lneq. (5.6).

Repeating the procedure with different

q rapidly gives an exterior approach of the

boundary of K and upper bounds tor the extreme loads (Fig. 11 ).

Furthermore it was proved, with some complementary mathematical assumptions,
that the exact dual definition of K is given by Form. (5.6), meaning that K can be
generated by applying Form. (5.6) to all k.a. virtual velocity fields [44-46].

Figure 11 - Kinematical approach from outside.

21

Yield Design: A Survey of the Theory

5.3.2.- Relevant velocity fields.

For lneq. (5.6) to give a non trivial result the k.a. virtual velocity field must be
chosen in such a way that :
- the position of the external forces be non zero
- the maximum resisting power be finite.
The first condition does not need any comment since it meets common sense. The

...

latter means that U should be selected in order to obtain finite values for the
corresponding x functions (cf. para. 5.2.2.).
Such virtual velocity fields will be said relevant. It must be emphasized that this
concept comes straight from the dual formulation of the strength condition and has
nothing to do with any flow rule or other constitutive assumption regarding the way the
material deforms or collapses when the strength condition is saturated.

5.4.- Some Classical strength criteria and corresponding

tr

functions.

5.4.1.- Definition of G(x) through a strength criterion.

As a rule, the strength domain G(x) is characterized, from the practical point of
view, by the data of a strength criterion through a scalar function f of a.(x), which
depends explicitly on x as does G(x), such that :
f(x, !:!.(X}) < 0
f(x, a..(x)) = 0
f(x, a_(x)) > 0

Q.(x)
a.(x)
a.(x)

<=>
:)
:)

interior to G(x) ,
on the boundary of G(x)
exterior to G(x) ,

(5.11)

For practical applications, function f(x, ) is usually chosen a convex, continuous,

with piecewise continuous derivatives, function of a.(x). it is clear anyhow that such a
function f(x, ) for a given G(x) is not unique.
With this definition of G(x), Eq. (5.2) may now be written, with simplified
notations, in the form :
x(x, 2.) = Sup{a.: g
l!

I f(x, a.)$ o}

(5.12)

Conversely, the data of function x(x, ) as the dual definition of G(x) provides an
expression for function f(x, ) :
f(x,a_)

S~P{.!l:.d.-x(!t) I tr.d.

=1}

(5.13)

this expression may not be the most popular one for the considered G(x) as regards
practical applications.

22

J.

Salen~on

Staying within the framework of classical continuum mechanics some usual

strength criteria and corresponding x functions for isotropic materials will now be
presented, together with the strength criteria and x functions of the most commonly
encountered isotropic interfaces. In order to simplify the expressions, the dependence
on

....

will not be explicited and the notation V will be used for (U].

5.4.2.- Von Mises's strength criterion and n functions.

f(.a.)

Ytr 1,.2/2 - k

(5.14)

where 1. is the deviatoric part of tensor

7t(d.}

thence:

if tr d.~ 0

00

x(d.) = k V2tr

.a..

d.,2

x(n, V)

+oo

1t(n, V)

k lVI

(5.15a)

if trd..=O

if V.n~O
if V . n =

(5.15b)

5.4.3.- Tresca's strength criterion and n functions.

f(SI.) = .. Sup
I,)=

1, 2, 3,

(5.16)

{oi- Oj- oo}

where Oi, oj denote the principal stresses.

1t(d.} = + 00
x(d.) = Oo (ld1l + ld2l + ld31)
2
where the di's are the principal components of

1t(n, V)
1t(n, V)

+ oo
Oo lVI

if tr d.~ 0

(5.17a)

if tr d.= 0

d. ; thence :

if V. n ~ 0
if V. n

=0

(5.17b)

5.4.4.- Coulomb's strength criterion and n functions.

f(.a.)

= .. Sup
1,)

=1, 2, 3,

{oi(1 + sin<p)- Oj(1 - sin<p)- 2 C cos<p}

1t(d.} = + 00
C cot<p tr d.
1t(d.}

if
if

tr d. < (ld1l + ld2l + ld31) sin<p

tr d.~ (ld1l + ld2l + ld31) sin<p

(5.18)
,

(5.19a)

thence

1t(n, V) = + oo
1t(n, V) = C cot<p V . n

if V . n < lVI sin<p

if V . n ~ lVI sin<p

(5.19b)

23

Yield Design: A Survey of the Theory

It will be observed that these expressions are different from the homologous
formulae given by Chen [19] which are only valid in the case when the inequalities in
(5.19} are saturated.

5.4.5. - Drucker-Prager's criterion and n functions.

f(Q.) = Vtr 1.,2 /2 -

(5.20)

h (

7t (.d.) = +oo
C cotq> tr d.

7t(.d.)

3 sincp
(C cotq>- tr SII3)
3+sin 2 q>)

if tr d.<

2sin 2cp (3tr

. 2 cp
3 +Sin

if tr

2sin 2 cp (3tr .d.2

. 2q>
3 +Sin

d.~

.d.2

(tr .d.) 2 )

.d.) 2 )

(5.21 a)
(tr

thence:
1t(n, V) = + oo
1t(n, V) = C cotq> V . n

if V . n < lVI sinq>

if V . n ~ IV I sinq>

(5.21 b)

5.4.6.- Tresca's strength criterion with T-tension cut off and n functions .
f (SI.)

. . Sup

7t(d.)

+oo

I, J = 1, 2, 3,

{cri- cri- Oo , Oi-

T}

(5.22)

if tr d.< 0
if tr d.~ 0

7t (.d.)

(5.23a)

thence:
1t(n, V)

+oo

7t(n, V)

0o

lVI + (T - !'!2.) V . n
2

if V.n<O
if V . n ~ 0

(5.23b)

When the material under consideration does not withstand any tensile stress the
value of T is set equal to nought.

5.4.7.- Coulomb's strength criterion with T -tension cut off and n functions.
f(Q.)

{ Oi( 1 + sinq>) = ..J =Sup

1, 2, 3,
I,

oi(1 - sincp) - 2 C cosq>, Oi- T}

(5.24)

J.

24

1t(sl}

+oo

1t(d.)

C coscp- T sincp
1 - sin<p

(ld11 + ld21 + ld31) + T- C coscp tr d.

1 - sin<p
if tr

d.~

Salen~ton

(5.25a)

(ld1l + ld2l + ld31) sin<p

thence:
7t(n, V)

1t(n, V)

+ oo

if V . n < lVI sin<p

C coscp- T sincp lVI + T-Ccoscp V.n

1 - sin<p
1 - sin<p
if V . n

(5.25b)

lVI sin<p

5.4.8.- Smooth or frictionless interface.

Vector T denotes the stress vector, with components (a, t), acting on the interface
with outward normal n. Vector V stands for the velocity jump in the interface.
f(T)

1t
1t

(5.26)

Sup (a, t)

(V) = + oo
(V) = 0

ifV.n<O,
if V.n~O

(5.27)

5.4.9.- Isotropic interface with Tresca's friction and no resistance to traction.

f(T) =Sup

{cr. t - Ci)

(5.28)

where Ci denotes the shear strength of the interface.

1t (V) = + oo
1t (V) = Ci Vt

if V.n<O
if V.n~o

(5.29)

where Vt is the modulus of the component of V in the interface.

5.4.10.- Isotropic interface with Coulomb's dry friction.

f(T)

= t

-a tan <pi

(5.30)

where <pi denotes the friction angle of the interface.

1t (V)
1t (V)

= + oo
=

if Vn < V1 tan<pi
if Vn ~ Vt tanq>i

(5.31)

25

Yield Design: A Survey of the Theory

5.5.- Implementation of the kinematic approach from outside.

5.5. 1.- Classical applications.

Usually the practical implementation of the kinematic approach from outside makes
use of simple relevant velocity fields depending on a reduced number of parameters. The

....

optimization of the obtained results is gained through the minimization of Pmr(U) with
respect to these parameters while keeping q(U) constant. Examples of such classically
used velocity fields may be given now.
Piecewise rigid body motion virtual velocity fields where one rigid block is moving
and the rest of the structure is motionless, or more elaborated virtual "mechanisrns"
where several rigid blocks move with respect to one another, will be encountered
(Fig. 12). In both cases the virtual velocity jumps will be governed by the condition
that n(x, n(x), [U(x)]) remains finite as explained in para. (5.3.2)
- Tresca's or von Mlses' strength criterion.

For instance, when dealing with a material whose strength criterion is of the
Tresca or von Mises types, it comes out from Eq. (5.15) and (5.17) that such k.a.
relevant virtual velocity fields will only exhibit purely tangental velocity jumps ;
examples are shown in Fig. 12:

,....

[U(x)] . n(x)

=0

(5.32)

Figure 12 - Indentation of a half space by a rigid plate : relevant rigid body motion
virtual velocity fields in the case of Tresca's or von Mises' strength criterion.

Relevant virtual velocity fields will also be constructed where the strain rate field
is non zero : they will be governed by the condition that n(x, d(x)) be finite.
Still taking the case of a Tresca or von Mises strength criterion as an example, the
corresponding k.a. relevant virtual velocity fields will induce no volume change as can
be seen from Eq. (5.15) and (5.17) :

....

tr .d.(x)

0 .

(5.33)

26

J.

Salen~on

Such fields as shear velocity fields will thus be relevant for those
materials.
- Coulomb's strength condition -

The case of a material with a strength criterion of Coulomb's type, must now be
considered for it will help emphasizing some essential points of the kinematic approach.
Eq. (5.19) gives the conditions to be satisfied by relevant virtual velocity fields :
A

""

tr .d.(x) ~ (ld 1 (x)l + ld2(x)l + ld3(x)l) sincp

(5.34)

which establishes a minimum value for the virtual dilatancy ;

"'

"'
[U(x)]. n(x) ~ I[U(x)]l sincp

(5.35)

which shows that the velocity jump, if non zero, cannot be purely tangential : [U(x)]
must be inclined over the jump surface I: by an angle ~ at least equal to cp and be directed
outwards, corresponding to a virtual separation between both sides of the jump surface

(Fig.13).

~~--~------------~~

...

Figure 13- The velocity discontinuity [U] of a relevant k.a. virtual velocity field
in the case of Coulomb's strength criterion.

It must be underlined once again that conditions (5.34, 5.35) are imposed on
virtual velocity fields and do not claim any significance from the point of view of any

flow rule, ... ; this is often misunderstood. Moreover it must be recalled that any
velocity jump is allowable for a virtual velocity field : velocity jump surfaces need not
be slip surfaces where the velocity jump is tangential. Therefore conditions (5.35) and

(5.32) are exactly on the same status, proceeding only from the strength condition of
the considered material. The latter point has often proved a difficulty in the
understanding of the kinematic approach from outside, in soil mechanics for instance,
through it is the key of the efficiency of the method.

Yield Design: A Survey of the Theory

27

Obviously from Eq. (5.21, 5.23, 5.25), what has just been said for Coulomb's
strength criterion will apply to the strength criteria considered in para. (5.4.5. 5.4.7.).
- Plane strain relevant k.a. virtual velocity fields -

When the problem under consideration accepts plane strain velocity fields as
kinematically admissible the simplification due to passing from dimension 3 to
dimension 2 makes it possible to construct more elaborate relevant k.a. virtual velocity
fields.
Considering first the case of Tresca's or von Mises' strength criterion (or any
criterion independent of tra), Eq. (5.33) applied to virtual k.a. velocity fields In plane
strain parallel to Oxy results in generating the relevant virtual velocity fields in the
following way.
Considering two families of mutually orthogonal curves (a- and

....

components Ua and

....

U~

....

,..

,..

,..

=0

along the a-lines

d.J~ + Uade = 0

along the ~-lines

where 9 is defined as 9 = (Ox, ea), ea and

~-lines

the

of U at the current point verify the differential equations [47]:

d.la- U~d9

and

~-lines),

at the current point, (ea,

ep

e~) =

{5.36)

being the unit vectors tangent to the a-

+ 7t/2 (Fig. 14).

For Coulomb's strength criterion, equations homologous to (5.36) are established

for the plane strain velocity fields which saturate lneq. (5.34) (cf. [48]).

Figure 14- Plane strain relevant virtual k.a. velocity fields for Tmsca's strength
criterion: Geiringer's equations [47].

28

J.

Salen~on

A classical example of such a plane strain relevant k.a. virtual velocity field is
recalled in Fig. 15 where volumes (1 ), (3), and (5) are given rigid body motions
while in volumes (2) and (4) the velocity is constant and orthoradial ; the plane and
cylindrical surfaces ABB'A', BCC'B', and CDD'C' and the symmetric ones are velocity
discontinuity surfaces.

Figure 15 -Indentation of a half space by a rigid plate: Prandtl's velocity field

in the case of Tresca's strength criterion.
It may be added that, due to the corresponding simplifications, many yield design
problems are often treated within the plane strain assumption, at least for a first
study. A detailed presentation as to the significance of this treatment within the
framework of yield design may be found in [36] and many interesting solutions are
available in numerous textbooks.

5.5.2.- Numerical implementation.

The efficiency of the methods which have just been sketched is highly improved
when analytical methods for the construction of k.a. relevant virtual velocity fields are
combined with numerical minimization procedures. This has been the basis of computer
codes which have been developed for instance for application to soil mechanics stability
analysis problems (e.g. [49 - 51]).
Many attempts have also been made during the two past decades aiming at using
numerical methods directly for the application of kinematic approach from outside.
Finite element methods have been considered for a straight forward application of the
theory [52 - 56], which has proved to present difficulties as regards the generation of
strictly relevant virtual velocity fields, and the minimization procedure. In many cases
the results so obtained have scarcely been as good as those gained through classical
procedures where discontinuous relevant virtual velocity fields can usually be
generated and explored more easily.

29

Yield Design: A Survey of the Theory

Recently, in the case of Tresca's strength criterion, a numerical method based upon
the generation of stream functions has been developed and seems quite efficient [57,

58].
6.- AN EXAMPLE OF THE YIELD DESIGN KINEMATIC APPROACH [59].

6.1.- Stability of a vertical cut for a purely cohesive soil.

The stability of the vertical cut introduced in Fig.7 is now considered, assuming the
homogeneous constituent soil to be purely cohesive, i.e. with a Tresca strength
criterion.
The relevant k.a. virtual velocity fields for such a material have already been
discussed in para. (5.5.1 ).
Applying the same arguments as in Section (4.3) it appears that the kinematic
approach for the problem can be restricted to a slice of arbitrary thickness D parallel
to plane Oxy with no resisting forces nor resisting power developed in the parallel
sections at both extremities. (A detailed presentation of this type of yield design
problems -plane strain yield design problems- appear in (36] where it is shown how
they can be studied through a two-dimensional yield design analysis).
For the slice of thickness D the following k.a. virtual velocity field can be retained
in accordance with Eq. (5.32, 5.33) : the prismatic volume OABB'A'O' is given a virtual
downwards translation motion with velocity V parallel to AB, while the rest of the soil
mass remains motionless (Fig.16).

Figure 16- Stability of a vertical cut in a purely cohesive soil: kinematic approach.
The power of the external forces performed in this virtual velocity field by the
gravity forces writes :

( 6. 1)

30

J.

Salen~;on

while the maximum resisting power being developed along plane ABB'A' in the
tangential velocity discontinuity reduces to :

....

Pmr(U) = CV h D/cos a

(6.2)

Thence through lneq. (5.4)

('Yh/C) + s 4/sin2a

(6.3)

Looking for the minimum of the second hand of this upper bound with respect to a
gives a= x/4 and

(yh/C)+

(6.4)

as the best upper estimate of ('Yh/C)* for this class of one parameter k.a. virtual
velocity fields.
The following comments can be made.
(a) Comparing the upper bound (6.4) with the value obtained in Section (4.3)

(lneq. 4.1 0) for this particular case where cp = 0 shows that both results are identical :

(yh/C)

is proved a necessary condition for the stability of the cut.

(b) This is not a mere coincidence!
As a matter of fact the kinematic reasoning performed here with a uniform
translation motion given to volume OABB'A'O' results in checking the global equilibrium
of that volume, as regards the resultant component in the direction of the motion, of the
gravity forces inn and resisting forces developed on ABB'A'.
Through the use of function x(n, [U]) here, the dual procedure in the kinematic
approach
- automatically finds out the axis along which the global equilibrium equation will
produce a non trivial necessary stability condition due to the expression of the strength
condition (here this relevant axis is AB),
- selects the distribution(s) of the resisting forces which most favour(s) the
global equilibrium of the volume so that external forces being not balanced under these
conditions will necessarily mean the instability of the cut.
(c) One may question about the admissibility of the considered velocity field since
the tip AA' of volume OABB'A'O' being given the velocity V parallel to AB would
penetrate the horizontal surface of the motionless mass of the rest of the soil.

31

Yield Design: A Survey of the Theory

Many arguments have been put forward in order to solve this apparent paradox
which occurs quite frequently in the applications of the kinematic approach (cf. Fig.12
and 15). For those which are relevant they appear adequate to the problem under
consideration with no versatility to be applied to similar cases, and do actually miss the
fundamental point.
The answer lies in the very significance of virtual k.a. velocity fields, already
recalled in para. (5.5.1 ), that are the test functions, the "mathematical tools" in the
dualization procedure. This means that they are piecewise continuous with piecewise
continuous derivatives, but no condition is imposed on the possible velocity
discontinuity. In the application of the kinematic approach, the "strength" of the
material, through the corresponding x(n, [U]) function, imposed the choice of
tangential velocity discontinuity on the plane AA'B'B for the value of x(n, [U]) to
remain finite. No such condition is to be found

to impose any constraint on Q along AA'

and this can be understood from the following : the three-dimensional continuum model
does not take lineal densities of forces into account (nor does the two-dimensional model
with punctual ones) and therefore the static analysis performed in Section (4.3) did
not consider any lineal resisting force at the vertex of volume (1) along AA', which is
consistent with no term appearing on AA' in the dual formulation.
(d) Another result of the kinematic approach is to answer the question of whether

the consideration of the moment equation of the global equilibrium of volume (1) in the
static analysis performed at Section (4.3) would have led to an additional limitation on
(yh/C) for the stability necessary condition.
The answer is negative and is apparent from the dual formulation. Considering the
moment equation corresponds, in the dual formulation, to a rigid body rotational motion
of volume OABB'A'O' inducing a velocity discontinuity across plane ABB'A' which would
not be tangential everywhere, unless the axis of rotation parallel to Oz be rejected at
infinity in the direction normal to AB which is the case already examined.
This illustrates the fact that, but for that particular case which leads to the final
result already obtained, the strength capacities of the soil will have no limiting
consequance on the balance of the moment equation of global equilibrium of volume (1 ).
Thus lneq. (6.4) expresses the best result to be derived from the consideration of
the global equilibrium of triangular prismatic volumes.

32

J.

Salen~on

(e) The same reasoning, through the kinematic approach as the dual formulation of
the static analysis, shows that circular cylindrical volumes may also be considered for
the "global equilibrium check" and produce significant results. In that case the relevant
equation which provides a necessary stability condition on (yh/C) turns out to be the
moment equation with respect to the geometrical axis of the cylinder (Fig. 17).

X:-h

Figure 17 - Stability of a vertical cut in purely cohesive soil : kinematic approach

with a rigid body rotational motion and "global equilibrium check".
The result obtained by looking for the strongest necessary condition over all
cylindrical volumes is classical [16, 17] and is given by
(yh/C)+

3.83 .

(6.5)

This upper estimate is better than condition (6.4) ; this is no surprise since the
analyses with triangular prismatic volumes are but particular cases of the present
ones.
(f) Finally it also appears from the dual formulation that no other cylindrical

volumes can be considered significantly for a static analysis through the "global
equilibrium check" : they would not result in any further constraint on (yh/C)+.
The upper bound resulting from lneq. (6.5) for (yh/C)+ has long been the best
available for this problem, despite numerous attempts where even sophisticated
numerical methods have been used. De Buhan, Dormieux and Maghous (60] have quite
recently produced a better result, obtained rigorously without heavy numerical
procedure, through an ingenious relevant virtual velocity field :
(yh/C)+

3.817 .

(6.6)

33

Yield Design: A Survey of the Theory

It is clear that the importance of this result will not be assessed from its
consequences as regards practical applications : the analysis of the stability of a
vertical cut is somewhat of a symbol where the "slip circle" has been baffling the
endeavours of researchers for decades, although it could be proved that it was not to
give the exact answer to the problem.

6.2.- Stability of a vertical cut for a soil exhibiting both cohesion and
friction.
Assuming now the strength condition of the constituent soil to be of Coulomb's type,
the following relevant k.a. virtual velocity fields will be used which satisfy lneq.
(5.34, 5.35) : the prismatic volume OABB'A'O' is given a virtual downward translation
motion with the velocity V inclined at an angle

to AB with <p s P s 1t- <p while the rest

of the soil mass remains motionless (Fig. 18).

Figure 18 - Stability of a vertical cut in a frictional and cohesive soils :

kinematic approach.
The power performed in this velocity field by the gravity forces becomes :

p)
(6.7)
2
while the maximum resisting power being developed along plane ABB'A' in the velocity
Qq (U)

= yh D tana V cos(a +

discontinuity is equal to

(6.8)
It follows through lneq. (5.4)
+
( yh )

2sinp
tan<p sina cos(a+P)

(6.9)

J.

34

where a and

Salen~on

are two parameters with the constraints

0

~a+ ~ ~

<p~~~'Tt-<p

7t/2
.

Minimizing the second hand of lneq. (6.9) with respect to

(6.1 0)
~

corresponds to
(6.11)

~=<p

for which lneq. (6.9) becomes

(yh/C( ~ 2 cos<p/sina cos(a+<p)

(6.12)

(to be compared with lneq. (4.8)).

Minimizing then with respect to a (as already done at Section 4.3) gives
(yh/Cf ~ 4 tan(1t/4+<p/2)

(6.13)

as the best estimate of (yh/C)+ one may achieve for this class of two parameters k.a.
virtual velocity fields.
This results also deserves some comments to complete those already made in the
simpler case of the purely cohesive soil.
(a) As in the case of a purely cohesive soil it is found that the upper estimate

derived from the kinematic approach with the virtual k.a. velocity fields of Fig. 18 is
equal to the result derived from the static analysis of Section (4.3) for the same
volume (1) in Fig. 9 dealing with the global equilibrium of volume (1) from the point
of view of the resulting force.
(b) One may also notice that the optimality condition ~

= <p

in the kinematic

approach means that, in the static analysis, the distributions of (a, -r) along plane
ABB'A' which most favour the global equilibrium of volume (1) correspond to the
equality

1-rl = C - a tan <p

with -r in direction AB

(which is quite evident here).

It also shows that the significant global equilibrium equation which is to provide
the stability constraint on (yh/C) is the resulting force equation in the direction at
angle cp to AB (the direction of V).
(c) The optimality condition ~

= <p found

in this particular case deserves being

looked at in a more general perspective since the constraints on [U] for the relevant
virtual k.a. velocity fields derived from lneq. (5.35) are weaker than for a purely

35

Yield Design: A Survey of the Theory

cohesive soil, namely

(6.14)
The proofs of the statements to appear hereafter can be found in the results given in
(36) : they rely on the kinematic dual formulation of the "global equilibrium check"
static analysis and can be generalized whatever the slope of the cut.
The idea would be to profit by the liberty left by condition (6.14) to perform the
kinematic approach using more general piecewise rigid body motion velocity fields with
any cross section for the cylindrical volume in motion. Such a rigid motion will be
either a rotation around any axis lz or a translation with any velocity V (Fig. 19).

Figure 19 - Stability of a vertical cut : more general kinematic approach

using piecewise rigid body motion velocity fields.
From the static analyis of view this would amount to the "global equilibrium check"
of a cylindrical volume (1) = OACC'A'O', with any cross section, being considered
thoroughly.
For the homogeneous vertical cut the results are as follows.
Starting from the kinematic approach it comes out that :
for a given axis lz defining a rotation with a velocity

(J)

(Fig. 20) the most

critical volume OACC'A'O', that is the volume leading to the strongest constraint on
(yh/C) + , is obtained when AC is an arc of a "cp" logspiral with pole I, so that the
induced velocity discontinuity be inclined at the angle

p =cp to the surface ACC'A' at each

point ; such a "cp" logspiral is defined in polar coordinates (p, 9) attached to I by the
expression p =Po exp(- a tan cp) ;

36

J.

Salen~on

as a particular case, when a translation is considered parallel to velocity V the

most critical volume OABB'A'O' is obtained when OAB is a triangle and V makes the angle

cp with AB in the outwards direction (Fig. 20).

Figure 20 - Stability of a vertical cut : most critical volumes for the kinematic
approaches by piecewise rigid body motion velocity fields.
Consequently, from the "global equilibrium check" point of view it follows that :
among all cylindrical volume (1) with any cross section being tested from the
"global equilibrium check" standpoint, the most critical are those with a cross section
defined by "cp" log spiral arc AC or a straight line AB (Fig. 21) ;

z~
X

Figure 21 - Stability of the vertical cut : most critical volume for the kinematic
"global equilibrium check".
for the first type, the significant equilibrium equation which leads to the
necessary constraint on (yh/C) is the moment equation with respect to lz where I is the

37

Yield Design: A Survey of the Theory

pole of the considered logspiral ;

for the second type, the significant equilibrium equation is the resulting force
equation in the direction inclined at the outward angle cp to AB in the plane Oxy (Fig.
21) .
The best result obtained through these equivalent analyses turns out approximately
to fit the formula [19] :
(yh/C( :s; 3.83 tan(7t/4+cp/2)

(6.15)

(d) the difference between Eq. (5.32) and (5.33) on one side, and lneq. (5.34)

and(5.35) on the other, have led to conclusions concerning the general conditions to be
fulfilled by relevant virtual k.a. velocity fields which might seem antinomic and make
it difficult to imagine the possibility of passing continuously from one case to the other
when cp

0.

This can be explained for instance for the relevant velocity discontinuities. For a
frictional and cohesive soil (cp ~ 0, C ~ 0) it has been explained that the relevant (U]
must belong to the cone drawn in Figure 22. Making cp tend to zero, the allowable cone

...

for [U] becomes wider and flattens down into the upper half space (Fig. 15). At the
same time it is seen from Eq. (5 .19) where C is a constant that, keeping

I[U) 1

constant, the value of 1t(n, [U]) tends to infinity when cp ~ 0 unless [U] belongs to the
very boundary of the cone. It follows that, cp tending to zero, the relevant [U] will tend
to be confined in the velocity jump surface.

(f)/E
8

~---------------~~

Figure 22- The relevant velocity discontinuities as cp ~ 0.

38

J. Salent;on

6.3.-

Comments.

The stability analysis of the vertical cut has been chosen in this presentation with
the purpose of giving an illustrative example of the kinematic approach being the dual
formulation of the static one. This point of view is quite obvious in the comparison
between Eq. (6.12) and Eq. (4.8) for instance.
Consequently, since only rigid motion virtual velocity fields have been used, this
example is not the best suited to enhance the possibilities of the kinematic approach
from outside, whose efficiency results from the variety of relevant k.a. virtual velocity
fields which can be used : they make it possible and easy, through the dual formulation,
to check the compatibility between equilibrium and resistance in a more elaborate way
than the mere global equilibrium of one or a few volumes. As a matter of fact the upper
bound (lneq. 6.6) obtained in [60] is derived from such a kinematic approach where
the virtual velocity field combines a rigid body motion velocity field and a shear
velocity field. A good example to illustrate this point of view would be the indentation of
a half space by a rigid plate (sketched in Fig. 12 and 15) with the classical results
recalled in [36], and more recent developments, such as those presented in [61] in the
case of an excentrated and inclined load on the plate acting on a purely cohesive soil with
no resistance to tensile stresses.
7.- CONCLUDING REMARKS.
7.1.-

The fundamental principles.

Concluding this survey of the Theory of Yield Design, the fundamental principles
will first be recoiled. The yield design theory relies on the necessary stability condition
that "Equilibrium and Resistance should be Compatible". From this follows the static

approach. Two papers by Hill [62, 63] may be referred to as early application of this
concept.
The so called "rigid-perfectly plastic model", which has unfortunately been
connected with this theory, is but an artefact without any physical reality nor
theoretical or practical utility ; its only effect has been confusing ideas and giving a
restrictive perspective of the possible field of applications of the yield design concept.
The kinematic approach is but the mathematical dualization of the static one. The
involved velocity fields are virtual velocity fields. They do not claim any actuality as
regards the true collapse modes of the considered system, even though experience seems

39

Yield Design: A Survey of the Theory

to show some similarity between observed collapse modes and "good" relevant k.a.
virtual velocity fields.

7.2. Possible fields of applications In civil engineering.

The short historical review at the beginning of this paper already gave a list of
various applications of the theory. A few words may now be added regarding recent or
possibly future developments in soil mechanics, taking advantage of the versatility of
the method:
stability analyses of seabed soils [64, 65], of tunnels [66, 67) ;
stability analyses assuming T or zero tension cut off strength criteria (para. 5.4.6
and 5.4.7) as first illustrated in the paper by Drucker [68], (e.g. [61]) ;
reinforced soil structures stability analyses through methods combining the yield
design and the homogenization theories [33-35], or through mixed modelling, that is
adopting a continuum mechanics model for the constitutive soil and a beam (or a strip)
model for the reinforcing inclusions : the interaction of the two models is mechanically
consistent and the strength criteria and

1t

functions are available ; the sound mechanical

bases make a clean sweep of all difficulties usually encountered in trying to adapt
"tradtional" stability analysis methods to these new circumstances.
Moreover, as already announced in [59], the same sound mechanical bases of the
theory make it the reliable way to build stability analysis methods and computer codes
which thoroughly comply with both the spirit and the letter of the Ultimate Limit State
Design (ULSD).
As a matter of fact the approach of safety which is conveyed in ULSD requires a
clear distinction to be made between the given loads on the considered structure and the

resistances which may be mobilized in the constituent materials (soil, reinforcement,

interfaces, ...). It then introduces partial safety factors to be applied : as multipliers on
the loads, greater than unity when the concerned load is unfavourable to stability,
lower than unity when it is favourable, they provide the design values of the loads ; as
dividers on the strength parameters, greater than unity, they provide the design values
of the strength parameters.
"According to the principles of Limit States Design, the design criterion is simply
to design for equilibrium in the design limit state of failure. The design criterion could

J. Salenr;on

40

be expressed in the following way :

Rd ~ sd

(7.1)

Sd is the design load effect... The design resistance effect Rd-" [69].
In order to take the adequacy of the design method into account, the design value of
ratio Rd/Sd in lneq. (6.16), denoted
rs = Rd/Sd

(7.2)

and called "Method coefficient", will not be compared to unity but to a reference value,
depending on the method used for the assessment of stability, which is usually greater
than unity.
This short presentation of ULSD clearly shows how the Yield Design Theory is fully
suited to this approach of safety which appeared as early as 1927 in Denmark and was
fostered by Brinch Hansen [70].
Contrary to many traditional methods used for stability analyses (e.g. the method of
slices, ... ) it does make the required clear distinction between loads and resistances, in
the sense that the calculation of Rd on the "strength" side of lneq. (7.1) does not, in
any way, involve the loads which only appear in Sd on the "load" side of lneq. (7.1)
It will be easily recognized that the fundamental principle of ULSD, as quoted from
[69] and expressed by lneq. (7.1) , does not differ from the fundamental principle of
the Yield Design Theory expressed in Formula (3.8), once the loads and the strength
parameters have been modified through their respective partial safety factors.
lneq. (5.4 or 5.6) and lneq. (7.1)

are not only similar but actually identical,

which means that the kinematic approach clearly and rigorously results in the upper
bound approach of rs .
The reproach has sometimes been made to ULSD that, since all quantities are
modified through partial safety factors, the design "mechanisms" might (or will) have
nothing in common with those which may actually occur in case of collapse. Though
questionable, for it may apply as well to the global safety factor approach, this remark
may be partially answered to as follows. It turns out that calculation methods and
computer codes based upon the Yield Design Theory are very efficient in the sense that
they prove to be usually less time consuming than many traditional methods ; added to

Yield Design: A Survey of the Theory

41

their rigorous mechanical basis, this makes it possible to perform parametric studies
which are certainly the best way to investigate the behaviour of a structure.

REFERENCES
1.

GALILEO G., Discorzi e dimostrazioni matematiche intomo a due nueve scienze,

Dialogo secundo, Leyden 1638.
2. COULOMB C.A., Essai sur une application des regles de Maximis et Minimis a
quelques problemes de statique relatifs
/'architecture, Mltmoire a I'Acadltmie
Royale des Sciences, Paris, 1n3.
3. HEYMAN J., The stone skeleton, Int. Jl. Solids and Structures, vol.2, (1966), n2,
249-279.
4. HEYMAN J., The safety of masonry arches, Int. Jl. Mech. Sc., vol.11, (1969),
363-385.
5. HEYMAN J., Coulomb's memoir on statics: an essay in the history of civil
engineering, Cambridge, G.B., 1972.
6. HEYMAN J., The estimation of the strength of masonry arches, Proc. A.S.C.E., Part
2, n69, (1980), 921-937.
7. HEYMAN J., The masonry arch, Ellis Horwood Ltd., John Wiley, 1982.
8. DELBECQ J.M., Analyse de Ia stabilite des voiites en ma"'nnerie de Charles Augustin
Coulomb nos jours, Ann. Ponts et Chaussees, no19, (1981), 36-43.
9. DELBECQ J.M., Analyse de Ia stabilite des voOtes en ma"'nnerie par le calcul Ia
rupture, J. Mec. Th. Appl., vol.1, (1982), n1, 91-121.
10. MERY E., Equilibre des voOtes en berceau, Ann. Ponts et Chaussltes, (1840), I,
50-70.
11 DURAND-CLAVE A., Stabilite des voOtes en ma~nnerie, Ann. Ponts et Chaussees,
(1867), I, 63-96.
12. DURAND-CLAVE A., Stabilite des voOtes et des arcs, Ann. Ponts et Chaussltes,
(1880), I, 416-440.
1 3. CULMANN K., Die graphische Statik, Zurich, 1866.
14. FELLENIUS W., Calculation of the Stability of Earth Dams, Trans. 2nd Cong. on
Large Dams, vol.4, Washington D.C., 1936.
15. BISHOP A.W., The use of the slip circle in the stability analysis of slopes, Proc.
Eur. Cont. Soil Mach. and Foundation Engineering, Stockholm, 1954.
16. TAYLOR D.W., Stability of earth slopes, Jl. Boston Soc. Civ. Eng., 24, (1937), 3,
337-386.
17. TAYLOR D.W., Fundamentals of soil mechanics, John Wiley, 1948.
18. RENDULIC L., Ein Beitrag zur Bestimmung der Gleitsicherheit, Dar Bauingenieur
n19/20, 1935
19. CHEN W.F., Limit analysis and soil plasticity, Elsevier, 1975.
20. MASSAU J., Memoire sur /'integration graphique des equations aux derivees
partielles ; chap. VI: Equilibre limite des terres sans cohesion, Ann. Ass. lng. Ecole
de Gand (1899). Editions du Centenaire, Comite National de Mltcanique, Bruxelles.
21. KOTTER F., Die Bestimmung des Druckes an gebrOmmeten BlattfiSchen, eine
Aufgabe aus des Lehre vom Erddruck, Berl. Akad. Bericht, (1903), 229.
22. KOTTER F., Uber dem Druck von Sand, Berl. Akad. Bericht, 1909.

42

J.

Salen~on

23. SOKOLOVSKI V.V., Theorie der Plastizizllt, VEB Verlag Technik, Berlin, 1955.
24. SOKOLOVSKI V.V., Statics of soil media, Butterworths, London, 1960, (after the
2nd edition, Moscow, 1954).
25. SOKOLOVSKI V.V., Statics of granular media, Pergamon Press, 1965.
26. BEREZANCEW B.G., Probleme de l'equilibre limite d'un milieu pulverulent en
symetrie axiale, Moscow, 1952.
27. HILL R., The mathematical theory of plasticity, Clarendon Press, Oxford (G.B.),
1950.
28. JOHANSEN K.W., Bruchmomente der kreuzweise bewehrten Platten, Memoires de
I'A.I.P.C., vol.1, (1932), 277-296, Zurich.
29. JOHANSEN K.W., Brudlinieteorier, Copenhague, 1952.
30. SAVE M. et MASSONNET Ch., Calcul plastique des constructions, Vol.2, 2eme ed., Ed.
CBLIA, Bruxelles, 1973.
31. de BUHAN P., SALENCON J., TALIERCIO A., Lower and upper bound estimates for the
macroscopic strength criteria of fiber composite materials, Proc. IUTAM Symp. on
Inelastic Deformation of Composite Materials, Troy N.Y., 29 may-1st june 1990,
563-580, Springer-Verlag, 1991.
32. de BUHAN P., TALIERCIO A., A homogenization approach to the yield strength of
composite materials. Eur. J. Mech., NSolids, 10, (1991), 129-154.
33. de BUHAN P., SALENCON J., SIAD L., Critere de resistance pour le materiau "terre
armee", C.R.Ac.Sc., Paris, t. 302, serie II, (1986), 337-381 .
.34. SIAD L., Analyse de stabilite des ouvrages en terre armee par une methode
d'homogeneisation, These Dr. Ecole Nationale des Ponts et Chaussees, Paris, 1987.
35. de BUHAN P., MANGIAVACCHI R., NOVA R., PELLEGRINI G., SALENCON J., Yield
design of reinforced earth walls by a homogenization method, Geotechnique, vol.39,
(1989), n2, 189-201.
36. SALENCON J., Calcul a Ia rupture et analyse limite, Presses de I'Ecole Nationale des
Ponts et Chaussees, Paris, 1983.
37. JEWELL R.A., Compatibility, serviceability and design factors for reinforced soil
walls, Proc. Int. Geotech. Symp. on Theory and Practice of Earth Reinforcement,
Fukuoka (Japan), Balkema publ., 1988, 611-616.
38. de BUHAN P., Approche fondamentale du calcul a Ia rupture des ouvrages en sols
renforces, Th. Dr. Sc., Paris, 1986.
39. BREZIS H., Operateurs maximaux monotones et semi-groupes de contractions dans
les espaces de Hilbert, North-Holland, Mathematic studies, 1973.
40. HALPHEN B., SALENCON J., Elasto-plasticite, Presses de I'Ecole Nationale des Ponts
et Chaussees, Paris, 1987.
41. DRUCKER D.C., PRAGER W., Soil mechanics and plastic analysis or limit design,
Quart. Appl. Math., 10, (1952), 157-165.
42. MOREAU J.J., Fonctionnelles convexes, Seminaire, 1966.
43. PRAGER W., Theorie generale des etats limites d'equilibre, J. Math. Pures Appl.,
34, (1955), 395-406.
44. NAYROLES B., Essai de theorie fonctionnelle des structures rigides plastiques
parfaites, J. Mec., 9, (1970), 491-506.
45. FREMOND M., FRIAA A., Analyse limite - Comparaison des methodes statique et
cinematique, C.R.Ac.Sc., Paris, t. 286, (1978), A, 107-110.

43

Yield Design: A Survey of the Theory

46. FRIM A., La loi de Norton-Hoff generalisee en plasticite et viscoplasticite, These

Dr. Sc., Univ. Pierre et Marie Curie, Paris, 1979.
4 7. GEl RINGER H., Fondements mathematiques de Ia theorie des corps plastiques

isotropes. Mem. Sci. Math., 86, Gauthier-Villars, Paris, 1937.

48. SALENCON J., La theorie des charges limites dans Ia resolution des problemes de
plasticite en deformation plane, These Dr. Sc., Univ. Paris, 1969.
49. ANTHOINE A., Mixed modelling of reinforced soils within the framework of the yield
design theory, Comp. Geotech., 7, (1989), 67-82.
50. ANTHOINE A., Une methode pour le dimensionnement Ia rupture des ouvrages en
sols renforces, Rev. Fr. G~otech., 50, (1990), 5-21.
51 . ANTHOINE A., SALENCON J., Une optimisation d'ouvrages en sols renforces, Proc. XII
Int. Conf. Soil Mech., Rio de Janeiro, Brasil, 1989, 1219-1220.
52. ANDERHEGGEN E. and KNOPFEL M., Finite Element Limit Analysis using Linear
Programming, Int. J. Solids & Structures, 8, (1972), 1413-1431.
53. FREMOND M. and SALENCON J., Limit analysis by finite element method, Proc. Symp.
on the Role of Plasticity in Soil Mechanics, Cambridge, UK, 1973, 297-308.
54. DELBECQ J.M., FREMOND M., PECKER A., and SALENCON J., Elements finis en
plasticite et visco-plasticite, J. M~canique Appliqu~e. 1, (1977), 3, 267-304.
55. TURGEMAN S., Contribution au calcul des charges limites en milieux isotropes et
orthotropes de revolution par une approche cinematique numerique, These Dr. Sc.,
Grenoble, 1983.
56. PASTOR J., TURGEMAN S., Approches numeriques des charges limites pour un
materiau orthotrope de revolution en deformation plane, K. M~ca. Th~or. Appl., 2,
(1983), 3, 393-416.
57. MAGHOUS S., Determination du critere de resistance macroscopique d'un materiau
heterogene a structure periodique- Approche numerique, Th. Dr. Ecole Natioanle des
Ponts et Chauss~es, Paris, 1991.
58. de BUHAN P. and MAGHOUS S., Une methode numerique pour Ia determination du
critere de resistance macroscopique de materiaux heterogenes a structure
periodique, C.R.Ac.Sc., Paris, 313, (1991 ), II, 983-988.
59. SALENCON J., An introduction to the yield design theory and its applications to soil
mechanics, Eur. J. Mech., A/Solids, 9, (1990), nos, 477-500.
60. de BUHAN P., DORMIEUX L., and MAGHOUS S., Private communication.
61 . SALENCON J. and PECKER A., Capacite portante de fondations superficielles sous
sollicitations inclinees et excentrees (etude theorique), Journ~e "Problemas
Scientifiques de l'lng~nieur", Hom mage a D. Radenkovic, Palaiseau (France, 16
janvier 1992.
62. HILL R., On the limits set by plastic yielding to the intensity of singularities of
stress, J. Mech. Phys. Solids, 2, (1954), 4, 278-285.
63. HILL R., The extremal stress-field concept, J. Mech. Phys. Solids, 14, (1966),
239-243.
64. DORMIEUX L., Stability of a purely cohesive seabed soil under wave loading,
G~otechnique, 38, (1988), 121-123.
65. DORMIEUX L., Influence de Ia houle sur Ia stabilite d'un massif sous-marin, These
Dr. Ecole Nationale des Ponts et Chaussees, Paris, 1989.

44

J.

Salen~on

66. LECA E., PANET M., Application du calcul a Ia rupture a Ia stabilite du front de taille
d'un tunnel, Rev. Fr. Geotech., 43, (1988), 5-19.
67. CHAMBON P., CORTE J.F., Stabilite du front de taille d'un tunnel en milieu frottant,
Rev. Fr. Geotech., 51, (1990), 51-69.
68. DRUCKER D.C., Limit analysis of two and three dimensional soil mechanics
problems, J. Mech. Phys. Solids, 1, (1953), no4, 217-226.
69. KREBS OVESEN N., General Report, Session 30: Codes and Standards, Proc. XII Int.
Cont. Soil Mech. Found. Eng., Rio de Janeiro, 1989, vo1.4.
70. BRINCH HANSEN J., Earth pressure calculation, The Danish Technical Press,
Copenhagen, 1953.